AI Keeps Getting Better. Why Is Its Writing Still So Bad? (greyenlightenment.com)

Despite noteworthy advancements in AI's capabilities, recent evaluations highlight persistent shortcomings in AI-generated writing. Prominent models like Opus and Astra still struggle to produce fluid, concise prose, often resulting in overly wordy and less effective communications. The ability of tools like Pangram to detect AI-written content has improved drastically, making it increasingly difficult for students to use AI for assignments without facing scrutiny. This suggests a narrowing gap between the enhancement of detection technologies and AI writing performance. Experts argue that AI fails to replicate the nuances and inherent "tics" of human writing and that its attempts to be precise often lead to bloated, redundant text. There may also be less commercial incentive to improve AI writing software compared to coding or math applications, which are more lucrative. The art of writing encompasses not just clarity of thought but also the intricate weaving of ideas, transitions, and style—elements where AI still falls short. As a result, the gap between AI's technical prowess and its ability to produce genuinely engaging prose remains wide, underscoring the complexity of human communication that AI has yet to master.